Merge remote-tracking branch 'origin/feat/nodegroups' into HEAD

Author: jovial

README.md

@@ -64,8 +64,10 @@ unique set of homogenous nodes:
       - `conf`: A string with the [resource specification](https://slurm.schedmd.com/slurm.conf.html#OPT_Gres_1) but requiring the format `<name>:<type>:<number>`, e.g. `gpu:A100:2`. Note the `type` is an arbitrary string.
       - `file`: Omit if `gres_autodetect` is set. A string with the [File](https://slurm.schedmd.com/gres.conf.html#OPT_File) (path to device(s)) for this resource, e.g. `/dev/nvidia[0-1]` for the above example.
     Note [GresTypes](https://slurm.schedmd.com/slurm.conf.html#OPT_GresTypes) must be set in `openhpc_config` if this is used.
+  * `features`: Optional. List of [Features](https://slurm.schedmd.com/slurm.conf.html#OPT_Features) strings.
   * `node_params`: Optional. Mapping of additional parameters and values for
   [node configuration](https://slurm.schedmd.com/slurm.conf.html#lbAE).
+  **NB:** Parameters which can be set via the keys above must not be included here.
 
   Each nodegroup will contain hosts from an Ansible inventory group named
   `{{ openhpc_cluster_name }}_{{ group_name}}`. Note that:
@@ -79,18 +81,23 @@ unique set of homogenous nodes:
     This is used to set `Sockets`, `CoresPerSocket`, `ThreadsPerCore` and
     optionally `RealMemory` for the nodegroup.
 
-`openhpc_partitions`: Optional, default `[]`. List of mappings, each defining a
+`openhpc_partitions`: Optional. List of mappings, each defining a
 partition. Each partition mapping may contain:
   * `name`: Required. Name of partition.
-  * `groups`: Optional. List of nodegroup names. If omitted, the partition name
-  is assumed to match a nodegroup name.
+  * `groups`: Optional. List of node group names. If omitted, the node group
+     with the same name as the partition is used.
   * `default`: Optional.  A boolean flag for whether this partion is the default.  Valid settings are `YES` and `NO`.
-  * `maxtime`: Optional.  A partition-specific time limit following the format of [slurm.conf](https://slurm.schedmd.com/slurm.conf.html) parameter `MaxTime`.  The default value is
-  given by `openhpc_job_maxtime`. The value should be quoted to avoid Ansible conversions.
+  * `maxtime`: Optional.  A partition-specific time limit overriding `openhpc_job_maxtime`.
   * `partition_params`: Optional. Mapping of additional parameters and values for
   [partition configuration](https://slurm.schedmd.com/slurm.conf.html#SECTION_PARTITION-CONFIGURATION).
+  **NB:** Parameters which can be set via the keys above must not be included here.
 
-`openhpc_job_maxtime`: Maximum job time limit, default `'60-0'` (60 days). See [slurm.conf](https://slurm.schedmd.com/slurm.conf.html) parameter `MaxTime` for format. The default is 60 days. The value should be quoted to avoid Ansible conversions.
+If this variable is not set one partition per nodegroup is created, with default
+partition configuration for each.
+
+`openhpc_job_maxtime`: Maximum job time limit, default `'60-0'` (60 days), see
+[slurm.conf:MaxTime](https://slurm.schedmd.com/slurm.conf.html#OPT_MaxTime).
+**NB:** This should be quoted to avoid Ansible conversions.
 
 `openhpc_cluster_name`: name of the cluster.
 
@@ -159,50 +166,121 @@ accessed (with facts gathering enabled) using `ansible_local.slurm`. As per the
 in mixed case are from from config files. Note the facts are only refreshed
 when this role is run.
 
-## Example Inventory
-
-And an Ansible inventory as this:
-
-    [openhpc_login]
-    openhpc-login-0 ansible_host=10.60.253.40 ansible_user=centos
+## Example
 
-    [openhpc_compute]
-    openhpc-compute-0 ansible_host=10.60.253.31 ansible_user=centos
-    openhpc-compute-1 ansible_host=10.60.253.32 ansible_user=centos
+### Simple
 
-    [cluster_login:children]
-    openhpc_login
+The following creates a cluster with a a single partition `compute`
+containing two nodes:
 
-    [cluster_control:children]
-    openhpc_login
+```ini
+# inventory/hosts:
+[hpc_login]
+cluster-login-0
 
-    [cluster_batch:children]
-    openhpc_compute
+[hpc_compute]
+cluster-compute-0
+cluster-compute-1
 
-## Example Playbooks
-
-To deploy, create a playbook which looks like this:
-
-    ---
-    - hosts:
-      - cluster_login
-      - cluster_control
-      - cluster_batch
-      become: yes
-      roles:
-        - role: openhpc
-          openhpc_enable:
-            control: "{{ inventory_hostname in groups['cluster_control'] }}"
-            batch: "{{ inventory_hostname in groups['cluster_batch'] }}"
-            runtime: true
-          openhpc_slurm_service_enabled: true
-          openhpc_slurm_control_host: "{{ groups['cluster_control'] | first }}"
-          openhpc_slurm_partitions:
-            - name: "compute"
-          openhpc_cluster_name: openhpc
-          openhpc_packages: []
-    ...
+[hpc_control]
+cluster-control
+```
 
+```yaml
+#playbook.yml
+---
+- hosts: all
+  become: yes
+  tasks:
+    - import_role:
+        name: stackhpc.openhpc
+      vars:
+        openhpc_cluster_name: hpc
+        openhpc_enable:
+          control: "{{ inventory_hostname in groups['cluster_control'] }}"
+          batch: "{{ inventory_hostname in groups['cluster_compute'] }}"
+          runtime: true
+        openhpc_slurm_control_host: "{{ groups['cluster_control'] | first }}"
+        openhpc_nodegroups:
+          - name: compute
+        openhpc_partitions:
+          - name: compute
 ---
+```
+
+### Multiple nodegroups
+
+This example shows how partitions can span multiple types of compute node.
+
+This example inventory describes three types of compute node (login and
+control nodes are omitted for brevity):
+
+```ini
+# inventory/hosts:
+...
+[hpc_general]
+# standard compute nodes
+cluster-general-0
+cluster-general-1
+
+[hpc_large]
+# large memory nodes
+cluster-largemem-0
+cluster-largemem-1
+
+[hpc_gpu]
+# GPU nodes
+cluster-a100-0
+cluster-a100-1
+...
+```
+
+Firstly the `openhpc_nodegroups` is set to capture these inventory groups and
+apply any node-level parameters - in this case the `largemem` nodes have
+2x cores reserved for some reason, and GRES is configured for the GPU nodes:
+
+```yaml
+openhpc_cluster_name: hpc
+openhpc_nodegroups:
+  - name: general
+  - name: large
+    node_params:
+      CoreSpecCount: 2
+  - name: gpu
+    gres:
+      - conf: gpu:A100:2
+        file: /dev/nvidia[0-1]
+```
+
+Now two partitions can be configured - a default one with a short timelimit and
+no large memory nodes for testing jobs, and another with all hardware and longer
+job runtime for "production" jobs:
+
+```yaml
+openhpc_partitions:
+  - name: test
+    groups:
+      - general
+      - gpu
+    maxtime: '1:0:0' # 1 hour
+    default: 'YES'
+  - name: general
+    groups:
+      - general
+      - large
+      - gpu
+    maxtime: '2-0' # 2 days
+    default: 'NO'
+```
+Users will select the partition using `--partition` argument and request nodes
+with appropriate memory or GPUs using the `--mem` and `--gres` or `--gpus*`
+options for `sbatch` or `srun`.
+
+Finally here some additional configuration must be provided for GRES:
+```yaml
+openhpc_config:
+  GresTypes:
+    -gpu
+```
 
 <b id="slurm_ver_footnote">1</b> Slurm 20.11 removed `accounting_storage/filetxt` as an option. This version of Slurm was introduced in OpenHPC v2.1 but the OpenHPC repos are common to all OpenHPC v2.x releases. [↩](#accounting_storage)

defaults/main.yml

@@ -4,7 +4,7 @@ openhpc_slurm_service_started: "{{ openhpc_slurm_service_enabled }}"
 openhpc_slurm_service:
 openhpc_slurm_control_host: "{{ inventory_hostname }}"
 #openhpc_slurm_control_host_address:
-openhpc_partitions: []
+openhpc_partitions: "{{ openhpc_nodegroups }}"
 openhpc_nodegroups: []
 openhpc_cluster_name:
 openhpc_packages:

molecule/test1/converge.yml

@@ -7,7 +7,7 @@
       batch: "{{ inventory_hostname in groups['testohpc_compute'] }}"
       runtime: true
     openhpc_slurm_control_host: "{{ groups['testohpc_login'] | first }}"
-    openhpc_slurm_partitions:
+    openhpc_nodegroups:
       - name: "compute"
     openhpc_cluster_name: testohpc
   tasks:

molecule/test10/converge.yml

@@ -7,7 +7,7 @@
       batch: "{{ inventory_hostname in groups['testohpc_compute'] }}"
       runtime: true
     openhpc_slurm_control_host: "{{ groups['testohpc_login'] | first }}"
-    openhpc_slurm_partitions:
+    openhpc_nodegroups:
       - name: "compute"
     openhpc_cluster_name: testohpc
     openhpc_slurm_configless: true

molecule/test10/verify.yml

@@ -29,7 +29,7 @@
           batch: "{{ inventory_hostname in groups['testohpc_compute'] }}"
           runtime: true
         openhpc_slurm_control_host: "{{ groups['testohpc_login'] | first }}"
-        openhpc_slurm_partitions:
+        openhpc_nodegroups:
           - name: "compute"
         openhpc_cluster_name: testohpc
         openhpc_slurm_configless: true

molecule/test11/converge.yml

@@ -11,7 +11,7 @@
           batch: "{{ inventory_hostname in groups['testohpc_compute'] }}"
           runtime: true
         openhpc_slurm_control_host: "{{ groups['testohpc_login'] | first }}"
-        openhpc_slurm_partitions:
+        openhpc_nodegroups:
           - name: "compute_orig"
         openhpc_cluster_name: testohpc
         openhpc_slurm_configless: true

molecule/test11/verify.yml

@@ -26,7 +26,7 @@
           batch: "{{ inventory_hostname in groups['testohpc_compute'] }}"
           runtime: true
         openhpc_slurm_control_host: "{{ groups['testohpc_login'] | first }}"
-        openhpc_slurm_partitions:
+        openhpc_nodegroups:
           - name: "compute_new"
         openhpc_cluster_name: testohpc
         openhpc_slurm_configless: true

molecule/test12/converge.yml

@@ -11,7 +11,7 @@
           batch: "{{ inventory_hostname in groups['testohpc_compute'] }}"
           runtime: true
         openhpc_slurm_control_host: "{{ groups['testohpc_login'] | first }}"
-        openhpc_slurm_partitions:
+        openhpc_nodegroups:
           - name: "compute"
         openhpc_cluster_name: testohpc
         openhpc_slurm_configless: true

molecule/test13/converge.yml

@@ -7,7 +7,7 @@
       batch: "{{ inventory_hostname in groups['testohpc_compute'] }}"
       runtime: true
     openhpc_slurm_control_host: "{{ groups['testohpc_control'] | first }}"
-    openhpc_slurm_partitions:
+    openhpc_nodegroups:
       - name: "compute"
     openhpc_cluster_name: testohpc
     openhpc_slurm_configless: true

molecule/test14/converge.yml

@@ -7,7 +7,7 @@
       batch: "{{ inventory_hostname in groups['testohpc_compute'] }}"
       runtime: true
     openhpc_slurm_control_host: "{{ groups['testohpc_login'] | first }}"
-    openhpc_slurm_partitions:
+    openhpc_nodegroups:
       - name: "compute"
         extra_nodes:
           # Need to specify IPs for the non-existent State=DOWN nodes, because otherwise even in this state slurmctld will exclude a node with no lookup information from the config.